AI data center concentration raises risks, with Bitcoin proof of work proposed as a security model

Concentrating AI infrastructure in large data centers can increase pressure on natural resources and expose networks to concentration risk. Distributing computing capacity and combining proof of work with technology that connects blockchains to external data could offer a way to strengthen data security.

In a November 9 interview with Anthony Pompliano, Tillman Holloway, CEO of ArchPublic, argued that Bitcoin’s proof-of-work network offers lessons for AI infrastructure design. He said distributing essential infrastructure more widely could reduce the burdens associated with concentrating it in a few locations.

Holloway said clusters of large data centers can strain natural resources and provoke political opposition in surrounding communities. He proposed allowing participants to contribute computing capacity much as owners of solar installations supply electricity to the grid. Household appliances could also contribute processing power, he suggested.

For data security, Holloway advocated proof of work tied to the consumption of electricity or another scarce natural resource. He argued that keeping ahead of rapidly developing AI threats would be difficult, and that security must cover both individual devices and large corporate databases. Phone-based checks that verify a person is present were one example he gave of useful protection at the device level.

He proposed combining proof of work with oracles, which bring external information onto blockchains. In his view, the combination could make a blockchain difficult to breach through brute-force attacks while allowing real-world data to enter securely in real time. He presented this as his outlook for how data protection could develop.

Holloway also argued that AI is lowering the effort required to analyze and extract value from collected data, making innovation accessible to more people. He said proficiency with AI could become a competitive advantage in employment and anticipated substantial innovation over the next 10 years in collecting, using and aggregating data, then turning it into useful services.

Electricity demand management was a practical application he highlighted. Holloway said grids must accommodate peak demand even though those peaks occur infrequently. Improving battery capacity and using AI to make immediate adjustments could free up grid capacity, he argued, bringing demand management to the forefront as AI and energy infrastructure develop together.